Photoreceptor-targeted gene delivery using intravitreally administered AAV vectors in dogs

Abstract

Delivery of therapeutic transgenes to retinal photoreceptors using adeno-associated virus (AAV) vectors has traditionally required subretinal injection. Recently, photoreceptor transduction efficiency following intravitreal injection (IVT) has improved in rodent models through use of capsid-mutant AAV vectors; but remains limited in large animal models. Thickness of the inner limiting membrane (ILM) in large animals is thought to impair retinal penetration by AAV. Our study compared two newly developed AAV vectors containing multiple capsid amino acid substitutions following IVT in dogs. The ability of two promoter constructs to restrict reporter transgene expression to photoreceptors was also evaluated. AAV vectors containing the interphotoreceptor-binding protein (IRBP) promoter drove expression exclusively in rod and cone photoreceptors, with transduction efficiencies of ~ 4% of cones and 2% of rods. Notably, in the central region containing the cone-rich visual streak, 15.6% of cones were transduced. Significant regional variation existed, with lower transduction efficiencies in the temporal regions of all eyes. This variation did not correlate with ILM thickness. Vectors carrying a cone-specific promoter failed to transduce a quantifiable percentage of cone photoreceptors. The newly developed AAV vectors containing the IRBP promoter were capable of producing photoreceptor-specific transgene expression following IVT in the dog.

INTRODUCTION

Adeno-associated virus (AAV) vectors have numerous advantageous properties that have been exploited for successful gene therapy including: the ability to transduce post-mitotic cell populations, induction of long-term transgene expression and low immunogenicity within human patients and animal models.^1,2 Subretinal injection is the most commonly utilized delivery route for targeting of AAV vectors to retinal photoreceptors.³ Major disadvantages of subretinal delivery include: the induced retinal detachment may worsen the underlying disease process the therapy is intended to treat, and the necessity of a trained surgeon with specialized equipment for routine injection success.^3–5 Interest in circumventing these disadvantages has driven the development of AAV vector variants which readily penetrate the retina and transduce photoreceptors following intravitreal injection (IVT). Unfortunately, IVT results in increased exposure of AAV particles to the immune system;⁶ and is more likely to produce off-target transgene expression in non-ocular tissues, including the brain.⁷ To circumvent these adverse effects, photoreceptor-specific promoters may be used, which have been shown to inhibit AAV-mediated transgene expression in off-target cell populations following subretinal, intravitreal and intravascular administration.^8–10

Recent advances in AAV vector development, including rational mutagenesis and the technique termed ‘directed evolution’, have made widespread outer retinal transduction possible in rodent models following IVT.^10,11 Alteration of the capsid amino acid constituents of AAV vectors can increase their overall efficiency. Substitution of specific capsid tyrosine residues for phenylalanine (Y-F) promotes evasion of intracellular proteasomal degradation pathways,¹² and decreases the affinity of vectors for heparan sulfate proteoglycan receptors on the inner limiting membrane (ILM) surface.¹³ Serine and threonine residue substitutions are thought to further diminish intracellular phosphorylation and ubiquitination leading to enhanced transport of AAV vector particles into the nucleus.¹⁴ An additional threonine-to-valine (T–V) substitution in an AAV2 vector containing four Y-F substitutions was recently shown to increase photoreceptor transduction in mice 3.5-fold over a vector containing the Y-F substitutions alone.¹¹

Fundamental differences in retinal structure and physiology make translation of results from rodent models to human subjects challenging, as various studies have demonstrated identical AAV vector constructs can have significantly attenuated responses when used in large animal models.^10,15,16 Dogs are an important preclinical model for the development of retinal gene therapy, as they exhibit inherited retinal diseases that mimic those observed in humans, often the result of a mutation in an analogous gene.¹⁷ They also possess similar retinal anatomy to humans, including a recently described fovea-like region of high cone photoreceptor density.¹⁸ Efficacy studies performed in canine models have provided the foundational justification for clinical trials to be performed in multiple human retinal dystrophies.^8,19–22

The primary aim of our study was to evaluate the photoreceptor transduction efficiency of two novel AAV2 vectors containing capsid amino acid substitutions following IVT in dogs. A green fluorescent protein (GFP) reporter transgene was driven by either the interphotoreceptor-binding protein (IRBP) promoter or guanine nucleotide-binding protein alpha transducing activity polypeptide 2 promoter with an IRBP enhancer (GNAT2/IRBP) to restrict expression to targeted rod and/or cone photoreceptors. Both promoters produce robust photoreceptor transduction following subretinal injection in dogs when packaged with AAV5 or AAV8 capsids.^8,23 As a secondary aim, we investigated the correlation of regional transduction efficiency variation with ILM thickness measurements.

RESULTS AND DISCUSSION

GFP expression and safety

Two novel capsid substituted vectors, AAV2 (quad Y-F+T–V) and AAV2 (quad Y-F), with GFP transgene expression driven by either the IRBP or GNAT/IRBP promoter, were delivered separately via IVT into the right or left eye of six dogs (Table 1) immediately adjacent to the retinal surface along the visual streak (Figure 1a). A total of 4 × 10¹¹ vector genomes were injected in a volume of 200 µl. At 6 and 8 weeks post-IVT, confocal scanning laser ophthalmoscopy (cSLO) showed the six eyes injected with IRBP vectors (AAV2 (quad Y-F+T–V) IRBP or AAV2 (quad Y-F) IRBP) had increased GFP fluorescence above background tapetal autofluorescence in regions surrounding the retinal vasculature (Figures 2a–d). Of the six eyes injected with GNAT2/IRBP vectors (AAV2 (quad Y-F+T–V) GNAT2/IRBP or AAV2 (quad Y-F) GNAT2/IRBP), none showed increased fluorescence on cSLO at either time point post-IVT. Increased fluorescence was not observed in either eye of the six dogs using fluorescent fundus camera imaging at any time point. Aside from expected mild post-operative intraocular inflammation, characterized by grade 1 out of 4 aqueous humor flare, (grade 1 = Tyndall effect just barely visible with slit-lamp magnification in a darkened exam room) in all eyes and mild settled hyphema in two of eight eyes, no adverse effects following IVT were noted in any of the six dogs. These changes were present at examination 24 h following injection, but had resolved by 48 h post-IVT. This degree of inflammation is typical as a response to IVT and subretinal dosing procedures in our laboratory, and does not appear to be a vector-directed reaction due to rapid resolution of clinical signs.

Table 1.

Intravitreal dosing arrangement and AAV vector descriptions

	Beagle 1	Beagle 2	Beagle 3	Mixed-breed dog 4	Mixed-breed dog 5	Mixed-breed dog 6
Left eye	AAV2(quad Y-F) IRBP	AAV2(quad Y-FV) IRBP	AAV2(quad Y-F) IRBP	AAV2(quad Y-FV) GNAT2/IRBP	AAV2(quad Y-F) GNAT2/IRBP	AAV2(quad Y-FV) GNAT2/IRBP
Right eye	AAV2(quad Y-FV) IRBP	AAV2(quad Y-F) IRBP	AAV2(quad Y-FV) IRBP	AAV2(quad Y-F) GNAT2/IRBP	AAV2(quad Y-FV) GNAT2/IRBP	AAV2(quad Y-F) GNAT2/IRBP

Abbreviations: F, phenylalanine; GNAT2, guanine nucleotide-binding protein alpha transducing activity polypeptide 2; IRBP, interphotoreceptor-binding protein; T, threonine; V, valine; Y, tyrosine.

Figure 1.

Photoreceptor transduction efficiency following IVT of AAV2 (quad Y-F+T-V) IRBP and AAV2 (quad Y-F) IRBP. (a) Schematic demonstrating the visual streak (lime green oval) over which the cannula was targeted during injection, and regions from which sagittal cryosections were collected (blue vertical bars). Numbers (1–9) approximate regions from which individual transduction efficiencies were determined. Transduction efficiency was determined by dividing the number of GFP-positive rod or cone photoreceptors by the total number of outer nuclear layer cells in retinal images taken through a × 40 microscope objective. (b) Overall (values determined from regions 1–9 combined) mean transduction efficiency for eyes injected with AAV2 (quad Y-F+T-V) IRBP or AAV2 (quad Y-F) IRBP. Dark gray bars represent mean transduction efficiencies for cones and light gray bars represent rods; line bars represent s.e.m. No statistical differences were observed when the two vectors were compared (cones: P =0.43; rods: P=0.42). (c) Regional mean transduction efficiency for eyes injected with AAV2 (quad Y-F+T-V) IRBP and AAV2 (quad Y-F) IRBP. Statistical comparison data for retinal regions are shown in Table 2.

Figure 2.

Localization of maximal photoreceptor transduction following IVT of AAV2 (quad Y-F+T-V) IRBP and AAV2 (quad Y-F) IRBP. Representative confocal scanning laser ophthalmoscopy images obtained at 6 weeks (a and b) and 8 weeks (c and d) following IVT demonstrate increased GFP fluorescence along the major retinal vasculature (solid arrows). Autofluorescence of the canine tapetum is responsible for the increased signal in the superior half of the images. Immunohistochemistry on cryosections shows a high level of rod and cone photoreceptor transduction directly underlying major retinal blood vessels for both AAV2 (quad Y-F) IRBP (e) and AAV2 (quad Y-F+T-V) IRBP (f). Scale bar, 50 µm. GCL, ganglion cell layer; hCAR, human cone arrestin; INL, inner nuclear layer; N, nasal; ONL, outer nuclear layer; RPE, retinal pigmented epithelium; T, temporal.

AAV vector retinal transduction efficiency

At 8 weeks post-IVT, sagittal eyecup cryosections were taken from all eyes through the optic nerve head, as well as 2–4mm nasal and temporal of the optic nerve head, and used for quantification of rod and cone photoreceptors expressing GFP. Overall, AAV2 (quad Y-F) IRBP and AAV2 (quad Y-F+T-V) IRBP vectors transduced 4.3 ± 0.8% and 4.6 ± 1.4% of cones, and 2.2 ± 0.3% and 2.0 ± 1.2% of rods, respectively (Figure 1b). These rates were not significantly different (cones: P = 0.43 and rods: P = 0.42). Up to 11-fold increases in transduction rates of photoreceptors were noted beneath retinal blood vessels, with up to 31% of cones and 25% of rods expressing GFP (Figures 2e and f). These regions were excluded from statistical analysis to avoid bias when comparing regions with limited retinal vasculature (such as the cone-rich area centralis and visual streak) with other retinal regions. The GNAT2/IRBP vectors did not transduce a sufficient number of cone photoreceptors to allow quantification or statistical comparison.

The major aim of our study was to evaluate the ability of novel AAV vectors to transduce photoreceptors in dogs following IVT. We found both IRBP vectors were capable of photoreceptor-specific transgene delivery following IVT, an original finding in a large animal model. Previous studies injecting wild-type AAV2 intravitreally in dogs and primates have reported absent or limited photoreceptor transduction.^15,24 Recently, capsid-mutant AAV2 vectors with ubiquitous promoters were reported to produce high levels of photoreceptor transduction beneath retinal vessels following IVT in dogs and primates.^10,16 It is promising that the IRBP vectors in our study also produced strong transduction in these regions, at similar efficiency to the AAV2 (quad Y-F) vector capsid paired with the strong ubiquitous smCBA promoter used for IVT by Mowat et al.¹⁶ Greater than 90% of photoreceptors were transduced by this vector following subretinal injection.¹⁶ It must be noted that the vector dose used for IVT in our study was roughly threefold higher. Although transduction efficiency was not quantified in a study reporting transgene expression in primate photoreceptors following IVT of a novel mutant AAV2 paired with the ubiquitous CMV promoter, confocal images of regions beneath retinal vessels appear to demonstrate comparable results to those found in our study.¹⁰ That vector was injected at a dose more than tenfold higher than our study.

Transduction efficiency in our study was lower than previously reported in mice using the same vector capsids paired with the strong smCBA promoter; however, the transgenes were packaged in a self-complimentary manner for the mouse study which increases transgene expression levels, whereas the vectors in our study had single-stranded genomes.¹¹ Single-stranded genomes allow larger transgenes to be delivered, making them more applicable for replacement of defective genes causing human retinal dystrophies.²⁵ Kay et al.¹¹ showed a significant increase in the number of photoreceptors transduced when AAV2 (quad Y-F+T-V) was compared with the AAV2 (quad Y-F) following IVT in mice. A similar effect was not observed in our canine model, as the vector transduction efficiencies were not statistically different in any region. The reason for the lack of difference between the efficiencies of the vectors in the canine is unclear; and could be attributed to differences in cell-surface receptors or intracellular trafficking at any level between the ILM and the outer nuclear layer. The primary receptor recognized by AAV2 is heparan sulfate proteoglycan, although additional co-receptors exist.¹ A recent study demonstrated that alterations to the capsid amino acid composition of AAV2 affects the vector affinity for these cell-surface receptors, thereby impacting IVT-mediated transduction.¹³ Species variation in the composition and density of these receptors on the ILM or other retinal cell membranes would theoretically produce differences in the ability of AAV vectors to transduce the photoreceptors following IVT. Differences in endosomal transport and capsid processing¹² of the vectors between species could also account for the differences we observed.

Our findings are preliminary for large animal models and further optimization of vector efficiency is required for future therapeutic use. A minimum photoreceptor transduction threshold likely exists that must be exceeded to obtain a clinically significant therapeutic response, but is likely to vary between different disease states and even individual animals, as has been shown previously.²⁰ Photoreceptor-specific AAV vectors capable of producing therapeutic rescue following IVT in rodent models of retinoschisis showed a lower level of transduction when injected into wild-type mice.²⁶ Although our initial transduction efficiency was limited in normal dogs, we are very interested to assess vector efficiency in a number of our canine models of retinal dystrophy. If required, further optimization techniques might include additional vector capsid amino acid substitutions or manipulation of the host ILM to potentiate retinal penetration.

AAV vector retinal tropism

As capsid engineering has advanced to improve transduction efficiency of next-generation AAV vectors, off-target expression of transgenes has emerged as a significant safety concern.^27,28 The IRBP vectors both drove GFP expression in LM-cones, S-cones and rods exclusively; without any off-target retinal transduction observed (Figure 3). This finding is in line with the report of Beltran et al.,⁸ in which use of the IRBP promoter limited transgene expression to canine rods and cones following subretinal injection. By limiting transgene expression in off-target tissues through the use of a cell-specific promoter, risk of developing a transgene-directed immune response is minimized.²⁹ In five of six eyes that received the GNAT2/IRBP vectors, limited GFP expression was present only within cone photoreceptors beneath the retinal blood vessels (Supplementary Figures 1a and b). In one of six eyes (dog 4, Table 1), off-target transduction occurred in the superior nasal retina following IVT of AAV2 (quad Y-F+T-V) GNAT2/IRBP; GFP expression was present within all retinal cellular layers at this site (Supplementary Figures 1c–f). When paired with an AAV5 vector, the GNAT2/IRBP promoter/enhancer combination produced transgene expression limited to cones and a small number of rods with no inner retinal expression in mice,³⁰ and in all cone subclasses exclusively in dogs,²³ following subretinal injection. It is unclear why the off-target expression within the inner retinal layers was only seen in one region of one eye in our study; one possibility is the affected region was exposed to a higher dose of vector as a result of the IVT procedure.

Figure 3.

Photoreceptor-specific transduction following IVT of AAV2 (quad Y-F) IRBP and AAV2 (quad Y-F+T-V) IRBP. Representative photomicrographs demonstrate GFP expression was limited to rod and cone photoreceptors 8 weeks post-IVT (a, d). Transduction of L/M- and S-cone photoreceptor subtypes is demonstrated (white arrows) in panels (b, e), and (c, f), respectively. Co-labeling with the L/M-opsin and S-opsin antibodies is implied by approximation of a GFP-positive inner segment with an opsin-labeled outer segment in this confocal microscopy section. Scale bar, 50 µm. GCL, ganglion cell layer; hCAR, human cone arrestin; INL, inner nuclear layer; ONL, outer nuclear layer; RPE, retinal pigmented epithelium.

Distribution of photoreceptor transduction

Variation in transduction rates produced by the IRBP vectors was observed between retinal regions (Figure 1c, Table 2), and a consistent pattern in the geographic distribution emerged during data analysis. The central and nasal visual streak regions (Figure 1a, regions 2 and 5) showed significantly higher (P < 0.05) transduction rates than temporal retinal regions (Figure 1a, regions 7–9). Notably, 15.6% of cones were transduced in the central region of the cone-rich visual streak. However, transduction efficiency was greatly attenuated in the temporal region containing the area centralis (Figure 1, region 8), a structure in dogs similar to the human macula.¹⁸ This variation in transduction efficiency could not be easily explained by procedural variables such as patient positioning or the dosing location, as the dogs were placed in dorsal recumbency. With the injections performed by a right-handed surgeon, this resulted in a temporal approach in the right eye and a nasal approach in the left eye. The injection device was intentionally moved across the region of the visual streak during vector administration to expose both the nasal and temporal retina.

Table 2.

Mean photoreceptor transduction and statistical comparison of regions

Region (illustrated in Figure 1a)	Percent cones transduced (± s.d.)	Regions P < 0.05	Percent rods transduced (± s.d.)	Regions P < 0.05
Nasal superior (1)	4.1 (±5.4)		2.5 (±2.0)	6
Nasal visual streak (2)	7.8 (±5.3)	6,7,8	4.3 (±3.4)	3,6,7,8
Nasal inferior (3)	3.8 (±3.4)	5	1.2 (±0.6)	2,5,6
Central superior (4)	5.9 (±6.5)		2.7 (±3.2)	6,7
Central visual streak (5)	15.6 (±5.9)	3,6,7,8	6.6 (±6.9)	3,6,7,8,9
Central inferior (6)	0.7 (±1.6)	2,5	0.3 (±0.3)	1,2,3,4,5
Temporal superior (7)	0.5 (±1.1)	2,5,8	0.6 (±0.7)	2,4,5
Temporal visual streak (8)	1.5 (±2.3)	2,5,7	0.5 (±0.5)	2,5
Temporal inferior (9)	0 (±0)	a	0.2 (±0.4)	5

^aRegion 9 cones could not be included in linear regression analysis due to value of 0.

It has been suggested that variation in ILM thickness may account for differences in transduction efficiency of AAV vectors delivered by IVT in canines and primates.^15,16 The ILM acts as a barrier to transport of not only AAV,^31–33 but also stem cells³⁴ and nanoparticles.^35,36 To investigate this hypothesis, retinal sections labeled with anti-laminin antibody were used to measure ILM thickness. Overall, there were no significant differences in ILM thickness detected between the retinal regions (Figure 4, P = 0.24). Linear correlation analysis showed there was no correlation between the ILM thickness and average photoreceptor transduction rates (cones: r²=0.17, rods: r² = 0.21; Figures 4d and e). This result suggests ILM thickness alone does not account for regional variation in retinal transgene expression in the dog, and some other property of the temporal retina or vitreous humor affects vector transduction efficiency. Both heparan sulfate proteoglycan and laminin receptors are reported to be present within the ILM of rodents.^31,33 It is possible a variation in the density of these receptors exists between retinal regions in the dog, which differentially sequesters AAV2 particles as they are transiting into the retina. This will be a future direction of investigation in our laboratory.

Figure 4.

Assessment of regional retinal inner limiting membrane thickness. Representative photomicrographs of paraffin section from eyecups of normal dogs labeled with anti-laminin antibody (a, b) demonstrates thickness of the ILM (open arrows); subjectively, ILM was thinner in regions overlying large retinal vessels (solid arrow). There was limited variability of ILM thickness (c) across retinal regions 1–9 as defined in Figure 1a. Linear correlation analysis showed no correlation of ILM thickness with percent transduction of either cone (d) or rod photoreceptors (e). GCL, ganglion cell layer; ILM, inner limiting membrane; INL, inner nuclear layer; ONL, outer nuclear layer.

Like previous studies, our results demonstrate increased retinal AAV transduction in regions adjacent to and beneath the retinal vasculature.^10,15,16,37 There is a stronger attachment of vitreous cortex to the retina in regions where the retinal vasculature exists.^38,39 The ILM is thinner over these regions in primates, and there is separation of Muller cell endfeet over the vessels resulting in the presence of pores within the ILM.⁴⁰ Vitreous humor is reported to extend through these pores, which may account for vector accumulation at these sites and increased transduction within these regions.⁴¹ These reported findings, in conjunction with our inability to link ILM thickness with outer retinal transduction efficiency, suggest that differences in composition of the ILM may alter AAV permeability to a greater degree than ILM thickness.

In conclusion, based on our primary aim, we demonstrated both AAV2 (quad Y-F) IRBP and AAV2 (quad Y-F+T-V) IRBP were able to produce photoreceptor-specific transgene expression following IVT in dogs. The addition of a T-V capsid substitution did not significantly increase transduction efficiency of the vector as observed previously in mice. AAV2 (quad Y-F+T-V) GNAT2/IRBP and AAV2 (quad Y-F) GNAT2/IRBP were considerably less efficient, and some off-target retinal GFP expression was observed. Following evaluation of our secondary aim, it appears ILM thickness is not significantly variable across the retina in dogs. ILM thickness alone, therefore, is unlikely to account for regional variations in vector transduction efficiency. Identification of a specific property of the ILM and/or vitreous humor that contributes to regional variation in AAV transduction efficiency would contribute to advancement in design of therapeutic vectors meant for IVT in human patients.

MATERIALS AND METHODS

Animals

Three purpose-bred 10-month-old Beagle dogs (Marshall BioResources, North Rose, NY, USA) were used for injection of the vectors containing the IRBP promoter; and three mixed-breed purpose-bred dogs (ages 9, 16 and 58 months) were used for injection of vectors containing the GNAT2/IRBP promoter. All dogs were socially housed with a 12-h light:dark cycle. Animal care was in compliance with the Association for Research in Vision and Ophthalmology statement for the Use of Animals in Ophthalmic and Vision Research, and all procedures were performed following approval by the Institutional Animal Care and Use Committee.

AAV vectors

Four recombinant AAV vector constructs containing capsid amino acid substitutions were manufactured and purified at the Retinal Gene Therapy Vector Lab, University of Florida College of Medicine using previously described methods.^11,42 One AAV2/2 vector was mutated for substitution of four surface-exposed capsid tyrosine residues with phenylalanine (Y272F, Y444F, Y500F and Y730F; referred to as AAV2 (quad Y-F)). A second AAV2/2 vector was mutated to include the same four tyrosine to phenylalanine substitutions, plus an additional substitution of a surface-exposed T–V (T491V; referred to as AAV2 (quad Y-F+T-V)). Two separate promoters were used: (i) the 1.3 kb human IRBP, reported to specifically target rod and cone photoreceptors following subretinal injection;⁸ and (ii) the 277- bp 5′-flanking sequence of the human guanine nucleotide-binding protein alpha transducing activity polypeptide 2 promoter coupled with the 214-bp IRBP enhancer (GNAT2/IRBP),⁴³ which targets cone photoreceptors following subretinal injection in mice, with a small number of rod photoreceptors transduced.³⁰ Both promoters drove expression of GFP. Each capsid type was used to separately package each of the two promoter/reporter gene constructs, yielding a total of four vectors (Table 1).

IVTs

Vectors were prepared for injection by diluting stock supplies to a titer of 2 × 10¹² vg ml^{− 1} using sterile balanced salt solution (Alcon Laboratories, Fort Worth, TX, USA). In all, 200 µl of vector solution was injected into the vitreous humor immediately anterior to the retinal surface in a transverse plane along the visual streak (Figure 1) using a RetinaJect injector (SurModics, Eden Prairie, MN, USA) as previously described.^16,44 Direct visualization of the injector tip in close proximity to its shadow on the tapetal retina allows solution to be injected immediately adjacent to the retinal surface. Previous work in our laboratory using retinoid therapy showed that injection adjacent to the retinal surface produced more consistent results; therefore, we adopted this method for IVT of AAV vectors to potentially reduce sequestration of vector particles within the vitreous humor.⁴⁴ Post-operative treatment included antibacterial and anti-inflammatory medications as previously described for IVT and subretinal injections in our laboratory.¹⁶ The vector dose was selected based on results from a preliminary safety study utilizing similar vectors containing the ubiquitous CBA promoter (unpublished data). Briefly: four eyes were administered 2 × 10¹² vg ml^{− 1} by IVT and showed increased GFP expression compared with four eyes administered 2 × 10¹¹ vg ml^{− 1}. No adverse effects were observed at either vector titer, so the higher titer from this safety study was selected.

Ophthalmic examinations and imaging

Post-IVT all dogs received regular ophthalmic examinations and fluorescence fundus imaging as previously described.¹⁶ cSLO utilizing 488 nm laser-induced fluorescence (cSLO, Spectralis, Heidelberg Engineering, Carlsbad, CA, USA) was performed at 6 and 8 weeks post-injection under general anesthesia.

Eyecup collection and sectioning

All dogs were killed 8 weeks post-injection with an intravenous injection of sodium pentobarbital (Fatal Plus, Vortech Pharmaceuticals, Dearborn, MI, USA). Eyes were immediately enucleated and processed as previously described.⁴⁵ Sagittal cryosections (14 µm thick), three sections per slide, were collected (Leica CM3050-S cryostat, Leica Microsystems, Buffalo Grove, IL, USA) from the central, nasal and temporal eye-cups (Figure 1a), and stored at − 20 °C.

Immunohistochemistry and cell quantification

Details of primary and secondary antibodies used for immunohistochemical labeling can be found in Supplementary Table 1. Representative images from the superior, central and inferior aspect of sagittal retinal cryosections collected from the central, nasal and temporal regions of each eyecup (Figure 1a) were captured using a confocal laser scanning microscope (Olympus FluoView fv1000 Confocal, Center Valley, PA, USA) at × 40 magnification. Slides labeled with cone arrestin (hCAR) and GFP antibodies were used for counting purposes with regions overlying retinal vasculature avoided. Identifying information was masked, images randomized and retinal cells counted by a single observer (RFB). For each image, separate counts were produced for: all DAPI stained nuclei within in the outer nuclear layer (ONL), GFP-labeled ONL nuclei, hCAR-labeled ONL nuclei and GFP-labeled ONL nuclei co-labeled with hCAR.

Evaluation of ILM thickness

Six globes collected from recently killed normal 3-month-old mixed-breed dogs and fixed in 2% paraformaldehyde were submitted to the Michigan State University Diagnostic Center for Population and Animal Health for labeling with an anti-laminin antibody. The globes were post-fixed for 24 h in 10% formaldehyde and then paraffin embedded. Sagittal retinal sections were obtained from each globe to match the sections evaluated for vector transduction efficiency. The labeled slides were imaged using white light microscopy (Nikon Eclipse 80i, Nikon Instruments, Melville, NY, USA) to obtain representative × 40 images from the nine retinal regions detailed in Figure 1. The thickness of the ILM from each region was measured using computer software (Adobe Photoshop CS4, San Jose, CA, USA).

Statistical analysis

Differences in overall retinal rod and cone photoreceptor transduction efficiency for each vector, as well as transduction efficiency between regions of the retina, were compared using generalized linear regression analysis with a Tweedie distribution (SPSS, IBM, Armonk, NY, USA). To clarify effects of significant categorical predictors, we applied a Bonferroni adjustment. Differences in ILM thickness between regions of the retina were compared using ANOVA (Excel, Microsoft, Redmond, WA, USA). Results were considered significant if P < 0.05. Linear correlation was used to compare the ILM thickness with the photoreceptor transduction rates by region (Excel, Microsoft). Data are displayed as mean ± s.d.